beckfield redes de politica mundial.pdf
TRANSCRIPT
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
1/51
1018 AJS Volume 115 Number 4 (January 2010): 101868
2010 by The University of Chicago. All rights reserved.
0002-9602/2010/11504-0002$10.00
The Social Structure of the World Polity1
Jason Beckfield
Harvard University
World polity research argues that modern states are shaped by em-beddedness in a network of international organizations, and yet thestructure of that network is rarely examined. This is surprising,given that world polity theory implies that the world polity shouldbe an increasingly dense, even, flat field of association. This article
describes the social structure of the world polity, using networkanalysis of the complete population of intergovernmental organi-zations as it has evolved since 1820. Analysis of the world politysstructure reveals growing fragmentation, driven by exclusive ratherthan universalist intergovernmental organizations. The world polityhas thus grown less cohesive, more fragmented, more heterogeneous,and less small worldly in its structure. This structure reflects arecent rise in the regionalization of the world polity.
Research on the possible consequences of globalization has overshadowed
research on the forms of globalization. This is particularly true for political
globalization, or the shifting reach of political power, authority and forms
of rule (Held et al. 1999, p. 32). We know a great deal about the asso-
ciations between political globalization and changes in the modern state.For instance, there is evidence that involvement in international orga-
1 I am grateful to Manish Nag and Violet Yebei for superb research assistance, to Paul
Ingram for sharing data, and to the AJS reviewers and editorial collective for con-
structive suggestions. For helpful comments and conversations, I also thank Art Alder-
son, Josef Bruderl, Larry Burmeister, Ron Burt, Likwang Chen, Rob Clark, MitchDuneier, Hartmut Esser, James Evans, Jeremy Freese, Claudia Geist, Dirk Jan-Koch,
Josh Klugman, Beate Kohler-Koch, Peter Kotzian, Ryon Lancaster, James Mahoney,
Matthew Mahutga, Paulette Lloyd, Patricia McManus, Ming-Chang Tsai, Jocelyn
Viterna, and Yang Yang. Earlier versions of this article were presented at the Universityof Chicago Workshop in Organizations and Markets and Workshop in Sociology and
Cultures of Globalization and the Sociology Colloquium at Northwestern University.
This research was supported by fellowships from Indiana University and researchfunds from the University of Chicago and Harvard University. Direct correspondence
to Jason Beckfield, Department of Sociology, Harvard University, 33 Kirkland Street,
Cambridge, Massachusetts 02138. E-mail: [email protected]
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
2/51
World Polity Social Structure
1019
nizations is associated with policy in the domains of human rights (Hafner-Burton and Tsutsui 2005), democracy (Wejnert 2005; Torfason and In-
gram 2010), environmental protection (Frank 1997, 1999; Schofer and
Hironaka 2005), same-sex sexual relations (Frank and McEneaney 1999),
womens suffrage (Ramirez, Soysal, and Shanahan 1997), education
(Meyer, Ramirez, and Soysal 1992; Bradley and Ramirez 1996; Schafer
1999; Schofer and Meyer 2005), population (Barrett and Tsui 1999), and
womens political representation (Paxton, Hughes, and Green 2006). Much
of this evidence comes from models inspired by world polity theory, an
institutionalist approach that explains the unexpectedly high and rising
level of isomorphism among states as a function of embeddedness in a
singular and universalist world polity (Boli and Thomas 1997; Meyer
et al. 1997), which is conceptualized as a network of states, societies, and
international organizations. The network of intergovernmental organi-zations (IGOs), in particular, has been shown to influence patterns of
democratization (Torfason and Ingram, forthcoming), international trade
(Ingram, Robinson, and Busch 2005), neoliberal restructuring (Henisz,
Zelner, and Guillen 2005), international conflict (Boehmer, Gartzke, and
Nordstrom 2004; Hafner-Burton and Montgomery 2006), and transna-
tional social movement organizations (Smith and Wiest 2005).
This article examines the structure of that network. Although debates
over the effects of the world polity are ongoing (Guillen 2001a, 2001b)
and critics highlight the role of power and inequality in the world polity
(Beckfield 2003), the depth of existing knowledge about the policy effects
and cultural content of the world polity contrasts starkly against our more
shallow understanding of its structure. This contrast is all the more sur-
prising in light of the many implications of world polity theory and other
approaches to political globalization for the structure of the world polity.
The world polity is said to be a world of Durkheimian and Simmelian
integration (Meyer et al. 1997, p. 175), a decentralized world (p. 164),
a unitary social system, increasingly integrated by networks (Boli and
Thomas 1997, p. 172), and a rapidly growing web of global links that
envelop the world without regard for local topography and conditions
(Boli, Loya, and Loftin 1999, p. 77). Resonating with arguments in the
popular press for globalization flattening the world (Friedman 2006),
world polity theory carries largely untested static and dynamic implica-
tions. Statically, the contemporary world polity should exhibit an even,
densely interconnected, decentralized social structure. Dynamically, if
there is fragmentation, centralization, and structural heterogeneity in thenetwork, these structural characteristics should be in decline.
To explore these and other structural implications and build on the
sociology of the world polity, I use data on the population of IGOs as it
has grown since the first IGO was established in the early 19th century.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
3/51
American Journal of Sociology
1020
IGOs are one type of social organization, along with international non-governmental organizations (INGOs), transnational corporations (TNCs),
and other civil-society organizations, in which world culture becomes
embedded (Boli and Thomas 1997, p. 172) and diffused through IGO-
to-state, IGO-to-IGO, and other kinds of ties among organizations (Boli
and Thomas 1999). IGOs, along with TNCs and states, have been char-
acterized as the dominant global actors in world culture (p. 173).2 I
analyze IGOs because the structure of the IGO field should correspond
most closely to the even field of association implied by world polity theory,
given that inequality in the number of IGO memberships per state has
decreased dramatically (Beckfield 2003). IGOs are also essential to the
world polity because IGO memberships appear to be practically com-
pulsory for states (Boli et al. 1999, p. 76), and a recent network turn
in international relations scholarship and globalization research has shownthat international organizationsand IGOs in particularmatter for a
range of political and economic outcomes (Boehmer et al. 2004; Henisz
et al. 2005; Ingram et al. 2005; Smith and Wiest 2005; Hafner-Burton
and Montgomery 2006; Torfason and Ingram, forthcoming). Still, this
network turn has yet to be taken to understand the essential structural
properties of the world polity itself. That is, many researchers argue that
the network structure matters, but this work has not yet considered the
structural implications of world polity theory or other approaches to po-
litical globalization.3
I examine 18202000 to show the complete evolution of the IGO field
that began in the early 1800s, well after the establishment of the West-
phalian system of sovereign nation-states in the 17th century, and con-
tinued through the Concert of Europe during the first half of the 19th
century, the major European wars of the late 19th century, early attempts
at global governance such as the League of Nations in 1919, the insti-
tutionalization of the United Nations system in 1945, decolonization in
the 1960s, and the collapse of the Soviet Union in the 1980s. Analyzing
the world polity as an evolving social network over this period reveals
2 Boli (2005, p. 384) identifies the World Trade Organization (WTO) as one such or-
ganization, noting that while hard-boiled politicking among competing states is surely
at work in this process, so too is global cultural construction. The WTOs rules and
agreements, globally applicable and reflective of universalistic principles, have thecharacter of world law that shapes the context of action for firms, states, and other
actors. It turns out that the WTO is one of the organizations at the very center of
the network of IGOs analyzed below.3 The analysis below follows previous research in considering a relation that forms
world polity networks: that of membership of states in IGOs. Other relations, such as
consultancies and other forms of direct IGO-IGO and state-state ties are not considered;see Boli and Thomas (1999) and Slaughter (2004) for discussions of these and other
networks.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
4/51
World Polity Social Structure
1021
that even in the context of a declining (and now low) level of heterogeneityin the number of IGO ties per state, there is a rising (and now high) level
of heterogeneity in the pattern of states IGO ties. This social structure
reflects, in part, place, as the world polity exhibits significant regionali-
zation along with globalization.4
SOCIAL STRUCTURE IN WORLD POLITY THEORY
In an influential programmatic statement of world polity theory, Meyer
et al. (1997) argue that many features of the contemporary nation-state
derive from worldwide models constructed and propagated through global
cultural and associational processes (p. 144).5 The authors use the imag-
inary example of a newly discovered island society to illustrate what hasalready happened to practically all of the societies of the world after their
discovery and incorporation into world society (p. 146). Focusing on the
associational process of integration into the world polity, the island society
would quickly come under the scrutiny of . . . international organiza-
tions, and its state and its people would be expected to join international
bodies (p. 165). As the island society grew more deeply integrated into
the network, old institutionalist and new institutionalist forces (Stinch-
combe 1997) would work to transform the island state into a legibly
modern state with globally legitimated organizational structures and prac-
tices. The state would seek legitimacy through international organizations,
while international organizations would certify the rational-legal authority
of the state and offer aid in the form of agents and material resources
(Meyer et al. 1997). Structurally, this process would create dense ties be-tween the state and international organizations, and policy scripts would
diffuse more easily from the world polity to the state as the conduits from
the world polity to the state grew in size, creating more and more re-
dundant and reinforcing connections.
What international organizations would the island join? For world
polity theory, the United Nations (UN) and its agencies (e.g., the Inter-
national Monetary Fund [IMF]; World Bank Group; UN Education, Sci-
ence, and Culture Organization; International Labour Organization; and
World Health Organization) represent the ideal-typical international or-
4 For a sociological conceptualization of place, see Gieryn (2000).5 Citation data from the Thomson Reuters Web of Science database indicate the sig-
nificant influence of Meyer et al. (1997): as of 2009, the article was more frequentlycited, at 402 citations, than any other article appearing in the last 12 years of the
American Journal of Sociology or the American Sociological Review.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
5/51
American Journal of Sociology
1022
ganizations.6
These organizations influence policy, distribute resources,and include nearly every state in the international system as members.
In the language of social network analysis, in joining the UN, the island
state, previously an isolated node in the network, would now have ties
to other states in the network. This network would be dense (every state
has a tie to every other state), decentralized (every state has the same
number of total ties), cohesive (states are close together in world polity
space), and clustered into one very small world (Wasserman and Faust
1994; Watts 1999). This two-mode, or affiliation, network (Breiger 1974)
could be represented as a bipartite graph (Faust 2005), in which the nodes
could be partitioned into a set of IGOs and a set of states.
Of course, the world polity also includes other kinds of international
organizations, including organizations that restrict membership by level
of economic development (e.g., the Organization for Economic Cooper-ation and Development [OECD]), geographic region (e.g., the Association
of Southeast Asian Nations), economic sector (e.g., the Organization of
Petroleum Exporting Countries), linguistic heritage (e.g., Francophonie
Institutionnelle), religion (e.g., Organization of the Islamic Conference),
or geopolitical alignment (e.g., the North American Treaty Organization).
This complicates the story.
Using the language of social network analysis, the above examples
highlight the dual (Breiger 1974) or bipartite character of the world
polity network.7 That is, the network formed by IGOs includes two types
of nodes, states and IGOs, making it a two-mode network that can be
represented as a bipartite graph. The first mode foregrounds the state
nodes; it is a network of states that are interlinked through memberships
in organizations. The second mode foregrounds the IGO nodes; it is a
network of organizations that are interlinked through their member states.
Although each mode is part of the same overall network, each will have
its own structure, and together, the networks form a bipartite network.
Understanding the structure of the world polity, therefore, requires an
examination of its dualitythe relational networks of states and orga-
nizations. To date, work on the structure of the world polity has opera-
tionalized involvement in the world polity as an attribute of states rather
than a relational structure (Wallace and Singer 1970; Jacobson, Reisinger,
and Mathers 1986; Shanks, Jacobson, and Kaplan 1996; Boli et al. 1999;
Beckfield 2003; Rey and Barkdull 2005), although work on international
6 The UN also has extensive formal and informal ties to INGOs, and the role of INGOs
in the UN has grown in the 1980s and 1990s (Pubantz 2005).7 Of course, states are linked directly to other states through bilateral relations, as are
IGOs.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
6/51
World Polity Social Structure
1023
conflict and policy diffusion has recently taken a network turn (Boehmeret al. 2004; Ingram et al. 2005).
While world polity theory suggests a densely interconnected global net-
work (a strong version of this hypothesis is a maximally dense network
in which all states and IGOs are connected), the theorys dynamic im-
plications are actually stronger. Nothing in world polity theory or its
empirical applications suggests that the world polity has always been
densely interconnected, only that it has become more so.8 Friedman makes
a parallel argument for economic globalization in The World Is Flat: I
know that the world is not flat. . . . I am certain, though, that the world
has been shrinking and flattening for some time now, and that process
has quickened dramatically (2006, p. 460). Returning to the evolution of
political globalization, the fact that world polity theory is fundamentally
a theory of change is reflected in the use of event history analysis andother techniques for longitudinal data in so many studies that test hy-
potheses drawn from the theory (Strang and Chang 1993; Ramirez et al.
1997; Boli and Thomas 1999; Frank, Hironaka, and Schofer 2000; Hafner-
Burton and Tsutsui 2005; Polillo and Guillen 2005; Schofer and Hironaka
2005; Wejnert 2005).
For instance, Frank et al. (2000) offer evidence for the argument that
the blueprints for nation-state involvement [in environmental policy] are
drawn in world society, from where they diffuse to individual countries
(p. 96). The assembly of evidence in their article follows a logic common
to much world polity research: the content of world culture in a given
domain is examined, and event history analysis is used to examine policy
change (in this example, indicators of change include the adoption of
environmental impact assessment legislation and the foundation of en-
vironmental ministries). The analysis shows that these policy changes are
driven, in part, by the extent to which countries have open conduits to
world society (p. 105), measured as memberships in IGOs and INGOs.
As in many other studies, Frank et al. show significant associations be-
tween ties to international organizations and policy change. The inter-
pretation of these associations is debated among sociologists (e.g., Buttel
2000), and it remains possible that unobserved heterogeneity, reverse cau-
sality, or omitted variables may bias the event history models. Indeed,
while it has sparked much research on policy isomorphism, world polity
theory remains controversial.
One current controversy surrounds the question of decoupling (Meyer
8 Change in the structure of the world polity is said to be reflected both in the increasing
connections among states and in the increasing connections among international or-ganizations (Boli 2005, p. 387). For the structure of the network, this implies increasing
density and decreasing centralization in both the interstate and the inter-IGO networks.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
7/51
American Journal of Sociology
1024
and Rowan 1977; Meyer et al. 1997, pp. 15456), or the frequent discon-nect between policy and practice. Buttel (2000), for one, argues that the
adoption of environmental policies may reflect mere window dressing
that has no impact on the environment itself. World polity researchers
have marshaled evidence to counter these critical claims, showing that
involvement in international organizations improves actual human rights
practice and environmental quality (Hafner-Burton and Tsutsui 2005;
Schofer and Hironaka 2005), but the debate continues in light of vast
inequalities among states in resources and power. Researchers in the world
polity tradition (and the new network research that assesses related hy-
potheses) have also elaborated more complex statistical models in an at-
tempt to control for some of the domestic factors that doubtless matter
for the creation, modification, and adoption of policy scripts (Buttel 2000).
Before traveling further down this road, it is informative to step backfrom this debate and consider the social structure of the world polity itself.
Research on the effects of integration into the world polity suggests an
increasingly densely interconnected, singular world polity, in which policy
scripts diffuse smoothly among organizations and states as there is less
and less friction in the world polity. These scripts diffuse more easily
to those states that have stronger connections to the world polity, and it
is argued that all states are increasingly embedded into the world polity.
As with the imaginary island society, states increasingly plug in to the
world polity, joining global organizations like the UN. World polity re-
search recognizes the tremendous growth in the population of interna-
tional organizations (Boli and Thomas 1997, 1999), but these organizations
are implicitly assumed to be global or at least increasingly global in scope,
forming a singular world polity rather than multiple (regional?) polities.9
If this assumption were true, and international organizations were in-
creasingly global in scope (i.e., increasingly akin to the UN), then this
would generate an increasingly dense world political structure in both
modes of the network. That is, the world polity would have both increas-
ingly dense interorganizational ties through states and increasingly dense
interstate ties through organizations. However, if this assumption were
false, and international organizations were becoming less global in scope
(i.e., increasingly akin to the Association of Southeast Asian Nations),
then this would generate a world polity that is increasingly fragmented
9
Of course, research in the tradition of world polity theory acknowledges that thereis a great deal of substantive heterogeneity among international organizations: the
world polity is divided into a wide range of sectors (Boli and Thomas 1997). Boli
(2005, p. 394) notes that globalized authority is highly fragmented and differentiated,and within each distinct sector or niche it is sometimes highly effective, sometimes
little more than symbolic. It is, nonetheless, very much on the upswing.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
8/51
World Polity Social Structure
1025
and uneven in its structure. Rather than a singular flat or even smallworld, the world polity would be best described as a fractured topology.
Would it mean anything for world polity theory if this assumption of
an increasingly densely integrated world political structure were unreal-
ized? To put the matter more sharply, it could be argued that assuming
a cohesive social structure is unnecessary for world polity research, given
that there are robust empirical associations between states ties to that
structure (whatever it might be) and national policy. It could be the plug-
ging in, and not the plugging in to what, that matters for states (cf.
Gartzke, Li, and Boehmer 2001; Ingram et al. 2005). This is one of the
empirical questions raised by conceptualizing and analyzing the world
polity in explicitly network terms.
Conceptually, the structure of the world polity should still matter for
understanding its effects, for at least five reasons. First, a more complexstructure might account for some of the anomalies of world polity research
(cf. Cole 2005 and Hafner-Burton and Tsutsui 2005). Second, accounting
for structure might enable a higher-resolution rendering of the mechanisms
through which the world polity affects states (e.g., regional organizations
might intervene in the process of policy diffusion). Third, the social struc-
ture of the world polity might offer a partial account of decoupling (e.g.,
practice may be more tightly coupled to policy in regional polities that
are more densely tied together). Fourth, if it is the intensity of involvement
in the world polity and not its social structure that matters for states,
then it could be that some third factor explains both involvement in
international organizations and the adoption of relatively progressive pol-
icies in the domains discussed above. Fifth, if worldwide models, or global
policy scripts, are generated in a world society of international orga-
nizations, those policy scripts may cohere better and diffuse more easily
among densely interconnected regional organizations (implying highly
structured heterogeneity among policy scripts). Indeed, studying the world
polity as a network could contribute to a better understanding of alter-
native processes of policy diffusion by getting closer to the proposed net-
work mechanisms of world polity theory (Simmons and Elkins 2005;
Valente 2005; Dobbin, Simmons, and Garrett 2007; Torfason and Ingram,
forthcoming).
These arguments suggest that understanding the social structure of the
world polity may carry important implications for world polity theory
and research. To date, work on the structure of the world polity, like the
research on the effects of the world polity, has operationalized involvementin the world polity as an attribute of states rather than embeddedness in
relational networks of states and organizations (Wallace and Singer 1970;
Jacobson et al. 1986; Shanks et al. 1996; Boli et al. 1999; Beckfield 2003;
Rey and Barkdull 2005). The social structure of the world polity itself is
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
9/51
American Journal of Sociology
1026
rarely studied.10
This is an important gap in our knowledge because verydifferent social structures result from a world in which states increase
their involvement in a growing number of global organizations like the
UN, compared to a world in which states increase their involvement in
a growing number of regional organizations like the European Union
(EU). The former world yields a flat, dense, even, cohesive social structure;
the latter, a rough, disintegrated, uneven, fragmented one.
ALTERNATIVE APPROACHES TO POLITICAL GLOBALIZATION
Other approaches to political globalization come from political science,
including neorealism (or structural realism) and neoliberalism (or liberal
institutionalism) from the international relations subfield (Gilpin 1975;
Waltz 1979; Keohane 1984; Moravcsik 1997), and sociology, including
world-systems theory and field-theoretic economic sociology (Boswell and
Chase-Dunn 2000; Fligstein and Stone Sweet 2002; Fligstein 2008).11 Al-
though it is not the goal of this article to present a definitive test of world
polity theory versus the alternative approaches, it is important to recognize
that world polity theory is not the lone approach to political globalization.
The different approaches tend to emphasize different explananda, and,
in some cases, their structural implications are similar, but there are points
of divergence that are noted below (indeed, one debate surrounds the
effectiveness of international organizations, which is not addressed by the
structural analysis in this article).
The key distinction between world polity theory and its alternatives
and the signal contribution of world polity theoryis that world politytheory accounts for increasing isomorphism among states by taking world
culture seriously. Emphasizing culture over organizational structure, the
argument is that states increasingly look alike because they are increas-
ingly embedded in a world polity (an associational structure) that expresses
a world culture (structures of meaning) that constitutes the state as an
10 Wallace (1975) provides an early and partial exception to this tendency by examining
the effective distance between dyads of states based on their common membership in
IGOs, and Kim and Barnett (2000) examine the network of IGOs in the international
telecommunications field. While the network structure of the world polity itself hastended not to be the object of empirical scrutiny, Ingram et al. (2005) and Hafner-
Burton and Montgomery (2006) show that IGO network ties are associated with in-
creased international trade and diminished interstate conflict among state dyads with
more dense IGO connections. Finally, Beckfield (2008) examines part of the networkat two time points: two highly visible subsets of IGOs.11 For an extended treatment of neorealist alternatives from the perspective of worldpolity theory, see Boli and Thomas (1997, pp. 17172, 1999, pp. 1519) and Meyer et
al. (1997, pp. 14648).
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
10/51
World Polity Social Structure
1027
actor, defining what it means to be a state and prescribing what statesdo.
The alternative approaches, however, highlight instead conflict, power,
and national (especially economic and military) interests. For instance, in
world-systems theory, the argument is that the world polity reflects and
reproduces preexisting structures of domination, as international orga-
nizations serve as boards of directors for ruling states (Boswell and
Chase-Dunn 2000, p. 238). Structurally, this implies a densely intercon-
nected world polity but one that is also highly uneven and centralized
around the dominant core actors, especially in the domain of economic
rule making. Such a pattern of ties would produce a star-shaped but tightly
interconnected network topology, with core states at the center. An equally
plausible implication of world-systems theory would be that both the core
and the periphery are closely interconnected, as peripheral states formand join exclusive IGOs as a way of representing their own interests
(Beckfield 2003); such a scenario would produce a core/periphery or a
global north/south fracture in the topology of the world polity. A different
pattern of fragmentation would be anticipated by the field-theoretic ap-
proach to economic sociology (Fligstein and Merand 2002; Fligstein and
Stone Sweet 2002), which implies that the topology of the IGO network
should depend on, in part, the pattern of international trade, which gen-
erates demands for international regulation and coordination through in-
stitutionalization, which itself then generates new social fields. Given that
political-economic integration has reached its most advanced expression
in the EU, the structure of the world polity should be highly regionalized
in Europe, especially the network formed by economic IGOs (Fligstein
and Merand 2002).
A strict neorealist view implies a sparsely interconnected world polity,
as states pursue their individual interests and avoid the binding obliga-
tions of IGOs (Waltz 1979). To the extent that IGOs exist, they should
exist in a few limited domains in which states interests are enhanced by
the coordination capacities of international organizations and regimes
(Krasner 1985). Such a view of the world polity implies a very fragmented
topology: sparsely interconnected, decentralized, not cohesive, and large
in the sense that the paths connecting dyads of states and IGOs would
tend to be longer rather than shorter. The neorealist depiction of the world
polity should be an especially apt characterization of the field of military/
political IGOs.
On this structural score at least, the implications of neoliberalism aresimilar to those of world polity theory. The neoliberal approach in the
international relations field does view IGOs as subject to states interests
(especially in the bargaining that results in IGO formation), but it also
views international organizations as effective and capable of aligning and
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
11/51
American Journal of Sociology
1028
coordinating states interests. Cooperation can increase, and states in-terests can be reshaped by the social structures within which they are
embedded (Moravcsik 1997). Structurally, then, neoliberals would antic-
ipate an increasingly densely integrated, increasingly decentralized, in-
creasingly small world polity. The economic domain of the world polity
should be more densely interconnected (and also more decentralized) than
the other domains (Keohane 1984).
If the UN is the paradigmatic IGO for world polity theory, then or-
ganizations like the OECD that restrict membership to developed coun-
tries and regional organizations like the EU are paradigmatic for alter-
native approaches. While the use of IGOs for interstate competition,
especially by noncore states to resist the liberal economic order, has been
debated (Krasner 1985), the structural implications have been explored
only as they apply to the level (not the pattern) of world polity involvement(Wallace and Singer 1970; Jacobson et al. 1986; Shanks et al. 1996; Boli
et al. 1999; Beckfield 2003).
Structural implications follow from the types of IGOs that are estab-
lished by states. For instance, the Group of 24 (G24) was founded in 1971
by 24 less developed countries to represent the interests of poor countries
in international financial matters and counterbalance organizations like
the Group of 7 (G7, now G8) industrialized countries. To see the structural
implications, imagine that the world polity consists of just the G24 and
the G8. The network formed would be bipartite, with two IGOs and 32
states. In turn, the bipartite network generates two one-mode networks:
a network (here, a dyad) of two IGOs and a network of 32 states. The
IGO dyadic network would be disconnected since no G24 member also
belongs to the G8. Likewise, the interstate network would also be dis-
connected, with all G24 states tied to all other G24 states and all G8
states tied to all other G8 states. If there were no global organizations
like the UN and all IGOs were formed on the basis of geographical or
other attributes, the world polity as a network would be disintegrated
instead of dense, fragmented instead of cohesive, and, given that resource-
rich states belong to more IGOs (Boli et al. 1999; Beckfield 2003), highly
centralized and structurally uneven.
Of course, the world polity is shaped by both inclusive organizations
like the UN and exclusive organizations like the EU. Thus, nearly all
states have at least one tie to nearly all other states (e.g., the UN forms
a tie between Germany and Bangladesh), although they may have a
greater number of ties to certain states than to others. And nearly everyIGO is likely to be connected to nearly every other IGO by at least one
common member state (e.g., Germany forms a tie from the EU to the
UN), but some IGOs may share more member states than others. This
suggests that in static terms, the world polity blends structural density
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
12/51
World Polity Social Structure
1029
with disintegration, decentralization with centralization, homogeneitywith heterogeneity, and cohesion with fragmentation. Given their theo-
retical relevance, it is essential to estimate these static properties. But the
dynamics matter more. This is because the theories in question are theories
of change. How has the structure evolved? Has it become more or less
dense, more or less centralized, more or less a flat field of association?
Consider once again the newly discovered island society. The island
state might pursue its interests on the global stage by forming strategic
alliances with other states and joining international organizations. It
would probably seek membership in global organizations like the UN,
thereby tying itself to nearly every other state in the international system,
but, assuming limited economic resources and a location in the Pacific
Ocean, it would probably also seek membership in organizations of poor
countries like the G24 and regional organizations like the South PacificCommunity and Pacific Island Forum. If the example of the island il-
lustrates the incorporation of all national states into the world polity
(Meyer et al. 1997, p. 146), then the world polity should exhibit some
degree of regionalization.
The degree of this potential regionalization of the world polity is
unclear. Although debate over regionalization informs studies of economic
globalization (Fligstein and Merand 2002; Kim and Shin 2002), scholar-
ship on political globalization tends to neglect the substantial role of place
and geography (Nierop 1989; OLaughlin and van der Wusten 1990; Van
der Wusten and Nierop 1990). World polity theory suggests that region-
alization should peak after World War II, after extensive universalistic
organizing of global IGOs encourages organizing at the regional level
(Boli and Thomas 1999, p. 31). Existing political-geographic research on
the question of regionalism in the world polity shows significant and
growing regional clustering, but this research is restricted to 195080
(Nierop 1989). Further inquiry is required to determine how much the
world polity resembles a world of regions (Katzenstein 2005).
DATA
To describe the structure of the world polity, I use newly available data
on the population of IGOs assembled and distributed as part of the Cor-
relates of War Project.12 Pevehouse et al. describe the data in their article
12 Pevehouse, Nordstrom, and Warnke (2004, p. 103) note that the broadest understand-
ing of what constitutes an IGO is that the organization (1) is a formal entity, (2) has
states as members, and (3) possesses a permanent secretariat or other indication ofinstitutionalization such as headquarters and/or permanent staff. . . . IGOs are dif-
ferentiated from nongovernmental organizations (NGOs) based on the fact that the
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
13/51
American Journal of Sociology
1030
in the journal Conflict Management and Peace Science (2004). The dataand codebook are available online at http://www.correlatesofwar.org/.
The data consist of binary matrices of IGOs and states, where ones
denote membership ties among IGOs and states in a given year, and zeros
indicate nonmembership.13 For the purposes of this article on the evolution
of social structure in the world polity, the key advantages of this data set
are its wide historical scope and its comprehensive inclusiveness of in-
dependent IGOs.14 An IGO is included in the data if it (1) includes at
least three member states, (2) holds regular meetings at least once per
decade, and (3) has a permanent secretariat and headquarters. Conversely,
an IGO is excluded if it is an emanation from another IGO (Pevehouse
et al. 2004). This is substantively relevant because many IGOs, such as
those of the UN Regional Seas Program, emanate from global IGOs and
thus represent globalization rather than regionalization. Information onIGOs and their members comes from multiple sources detailed in Wallace
and Singer (1970) and Pevehouse et al. (2004), including the Yearbook of
International Organizations, published by the Union of International As-
sociations (UIA). The UIA is the standard source for data on world polity
ties (Boli and Thomas 1997; Beckfield 2003; Cole 2005). The data set
includes information on the memberships of a total of 495 separate IGOs
that existed at some point since 1815.15 The IGOs in existence as of 2000
are listed in the appendix.
The states included are those that meet the criteria for membership in
the interstate system, as defined by the Correlates of War Project (2005).
The criteria restrict the list to entities that (1) prior to 1920, . . . have
population greater than 500,000 and have had diplomatic missions at or
above the rank of charge daffaires with Britain and France and (2)
after 1920, [have membership in] the United Nations or League of
Nations, or have population greater than 500,000 and receive diplomatic
latter organizations memberships are composed of individual persons, interest groups,
or businesses.13 Ingram et al. (2005) use these data in a study of international trade. Oneal and
Russett (2001) and Russett and Oneal (2001) employ an earlier version of the data set
in research on militarized international disputes.14 Restricting the sample of IGOs to a limited subset of highly visible and especially
influential organizations is an important next step. The social structure of the world
polity may vary according to the prominence of the organizations that are analyzed,with prominent IGOs more likely to be global IGOs. Elsewhere, I find that the sub-
stantive conclusions reached in this article are identical in an analysis restricted to
prominent IGOs (Beckfield 2008).15 For the purposes of describing the network, all IGOs are treated as structural equals,
which elides important differences among the organizations that populate this field.
Including all IGOs in the network arguably biases the results toward supporting worldpolity theory because introducing more heterogeneity into the network would likely
produce a more uneven structure.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
14/51
World Polity Social Structure
1031
missions from two major powers. This means that it is possible for thebipartite network to be disconnected because states need not be a member
of any IGO to be included in the data set. The equivalent of the island
society that enters the world polity de novo is thus observed in this data
set, as the bipartite network is disconnected (and island societies are iso-
lates) before 1960. Both the list of IGOs and the list of states are time
varying, according to the entries and exits of states and organizations in
the international system (for further details, also see Small and Singer
[1982]).
Consistent with the focus of this article on long-term structural change
in the world polity, the analysis uses data at 10-year intervals: 1820, 1830,
and so on through 2000. Given the two-mode, or bipartite, structure of
the network of IGOs and states, each year of data produces three matrices
for analysis: an asymmetrical, two-mode binary matrix where cell xij in-
dicates the membership status of state j in IGO i (1 or 0); a symmetrical,
one-mode valued matrix where cell xij counts the number of IGO mem-
berships shared by states i and j; and a second symmetrical, one-mode
valued matrix where cell xij counts the number of member states shared
by IGOs i and j. In other words, there is a network of IGOs and states,
a network of states tied through IGOs, and a network of IGOs tied through
states (direct ties between states, such as diplomatic missions and bilateral
treaties, and direct ties between IGOs, such as consultative statuses, are
omitted). Each network reveals distinct aspects of the social structure of
the world polity. The first two-mode network, which can be represented
as a bipartite graph, reveals the density of the world polity as a whole,
while the one-mode networks reveal the centralization (or lack thereof)around key nodes and structural heterogeneity (or lack thereof) among
positions in the network. The bipartite graph can then be analyzed for
the overall cohesion and clustering of the network. Finally, the network
of states can also be examined for evidence that states from the same
geographic region share more IGO ties than states from different regions.
Both the network of IGOs and the network of states can be analyzed as
a valued matrix, where the cells are counts of ties, or as a binary matrix,
where the cells take the value of one if there is at least one tie. For instance,
in the valued IGO network, the value of the tie between the EU and the
IMF in 2000 is 15 because all 15 EU member states in 2000 were also
members of the IMF. However, in the binary IGO network, the value of
this tie equals one. Analyzing both valued and binary networks allows for
the assessment of the strength of connections as well as the presence of
connections. It is important to examine each aspect of the social structure
of the world polity because there may be less structural unevenness in the
presence/absence of network ties than in the strength of ties. Moreover, the
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
15/51
American Journal of Sociology
1032
strength of connections, or volume of conduits to the world polity, is arguedto be the network mechanism for the diffusion of policy scripts.16
In sum, at each 10-year interval between 1820 and 2000, the original
data matrix is two mode and binary, where the ties are between IGOs
and states. The rows in this matrix are IGOs, and the columns are states.
From this data matrix derive five matrices that capture distinct aspects
of network structure: (1) a valued matrix where the rows and columns
are IGOs and the cells count ties between IGOs, (2) a binary matrix where
again the rows and columns are IGOs but the cells indicate the presence
or absence of a tie between IGOs, (3) a valued matrix where the rows
and columns are states and the cells count ties between states, (4) a binary
matrix where again the rows and columns are states but the cells indicate
the presence or absence of a tie between states, and (5) a bipartite sym-
metric graph where all IGOs and states appear in both the rows and thecolumns and the cells indicate IGO-state ties. These matrices generate
decennial snapshots of the structure of the world polity, in which both
the composition of IGOs and the composition of states vary with time as
the world polity evolves.
ANALYSIS
World polity research to date has emphasized cultural content over or-
ganizational structure, but the structural implications of the theory, which
are developed above, are clear: the network should be (increasingly)
densely interconnected, (increasingly) decentralized as states join IGOs in
similar patterns, (increasingly) equal in the extent to which states holdmemberships in IGOs, (increasingly) cohesive in topology as states become
more reachable to each other through shared IGO ties, (increasingly) small
worldly in global structure, and (decreasingly) regionalized. Following
these structural implications (and the implications of alternative theories
developed above), I use network analytic techniques to examine change
in several essential structural properties: density, centralization, hetero-
geneity, cohesion, and clustering.17 To examine regionalization in the world
16 Of course, even the valued matrix weights every IGO (and every state) equally in
the network analysis. The structure of the world polity could be more heterogeneousif IGOs were weighted differently (according to some measure of influence or resources),
but the analysis in this article is not designed to assess this conjecture.17
Following the Simmelian insight that triads make more stable groups than dyads,Moody and White (2003, p. 103) operationalize structural cohesion in a social network
as a function of connectivity: structural cohesion is defined as the minimum number
of actors who, if removed from a group, would disconnect the group. As with manygraph-theoretic properties, connectivity is a property of binary networks that does not
extend readily to valued networks such as those analyzed here (Wasserman and Faust
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
16/51
World Polity Social Structure
1033
polity, I calculate correlations between the observed network and a modelnetwork in which ties are based on region. Analyses were performed using
the programs Ucinet 6 (Borgatti, Everett, and Freeman 2002) and Pajek
(Batagelj and Mrvar 2007).
Density, a fundamental property of social networks, is calculated as the
percentage of possible ties in the network that are actually observed (Was-
serman and Faust 1994). The numerator is a simple count of ties, but the
denominator differs according to the modality of the network. In a one-
mode network (e.g., a network of just IGOs), the denominator is the
number of nodes in the network multiplied by the number of nodes minus
one (nodes in these networks cannot be tied to themselves). In a bipartite
network, ties can only be observed between actors in different modes
(here, between IGOs and states), so the relevant denominator is the num-
ber of actors in the first mode multiplied by the number of actors in thesecond mode (Borgatti and Everett 1997). Density is calculated only for
the binary networks and ranges from 0 (where no ties are observed) to 1
(where all possible ties are observed). It is important to note that network
density (or relative density) differs from population density (or absolute
density) in that network density measures the realization of possible ties,
not the volume of possible ties itself. That is, there is no question that
one indicator of world polity formation is the dramatic increase in the
number of IGOs and states in the systemor the increase in the absolute
density of world polity ties. This distinction is akin to that between an
increase in the population of a given neighborhoodand thus an increase
in the potential for interactionand the actual realization of ties among
people in that neighborhood.
The centralization of a network is the degree to which it resembles a
star network, in which one central node has ties to every other node
but the other nodes do not have ties among themselves. For instance, if
the UN shared a member state with all the other several hundred IGOs
that populated the network in 2000 but these other IGOs did not overlap
in their membership, the IGO network would be perfectly centralized.
Like density, centralization is a structural property of the network as a
whole. It is calculated as
C (n ) C (n )D D iC p ,D
(g 1)(g 2)
where represents the degree centrality of the most central node,C (n )D
represents the degree centrality of node i, and g represents the totalC (n )D inumber of nodes in the network (Wasserman and Faust 1994, p. 180).
1994, p. 76). Also, this measure of structural cohesion produces trivial results when
applied to the interstate network because it is disconnected until 1960.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
17/51
American Journal of Sociology
1034
Centralization is calculated for binary networks, and ranges from 0 (per-fectly decentralized) to 1 (perfectly centralized).18
While centralization captures one sense of structural heterogeneity or
unevenness, it is limited in that networks can have more than one central
node. For instance, some networks resemble a core/periphery structure,
with a set of tightly interconnected nodes and a peripheral set of sparsely
interconnected nodes (Borgatti and Everett 1999). In core/periphery struc-
tures, it is possible to assign a coreness score to each node that quantifies
how close it is to the dense core of the network. In a maximally dense
network, all nodes would have the same coreness score because every
node would belong to the dense core, and no sparse periphery would exist.
Thus, dispersion in coreness scores can be used to assess overall network-
level structural heterogeneity. I use the genetic algorithm developed by
Borgatti and Everett (1999) to assign coreness scores, and I use the Gini
coefficient to measure dispersion in the scores. The Gini coefficient is a
commonly used measure of dispersion (see Firebaugh [1999] for an ap-
plication) that varies from 0 (perfect homogeneity) to 1 (perfect hetero-
geneity).
To measure the cohesion of the IGO-state network, I use the bipartite
graph to calculate the average path length, diameter, and connectivity of
the graph (Wasserman and Faust 1994). A path is the smallest number
of ties (or links) between two nodesfor instance, if a policy script must
travel from the UN to APEC (Asia-Pacific Economic Cooperation) through
a shared member state to reach Taiwan, the UN-Taiwan path would be
three ties long, and the UN-APEC path would be two ties long. The
average path length is calculated over all dyads in the network. Thediameter of the graph is the maximum rather than the average path length.
Diameter is useful for comparing bipartite graphs because it measures
how far apart the two most distant nodes are. The third measure of
cohesion is connectivity: the minimum (node) connectivity of a graph is
a count of the number of nodes that would have to be removed from the
graph to disconnect it. To provide a baseline for comparison, I also cal-
culate these cohesion indexes for random networks (the random networks
18 Supplemental analysis indicates that using the binary network to calculate central-
ization, as is conventional given that centralization for binary networks is bounded
between zero and one and thus has a ready interpretation, results in understatementof the increased centralization of the IGO network and overstatement of the decreased
centralization in the state network. Using the valued networks, the increase in the
centralization of the inter-IGO network is more pronounced, while the decrease in thecentralization of the interstate network is less pronounced. Thus, the figures below
can be seen as conservative depictions of the trends in world polity centralization.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
18/51
World Polity Social Structure
1035
were generated by holding constant the size and tie distributions of thenetworks and creating random connections between nodes).19
I follow Uzzi and Spiro (2005) in calculating the clustering ratio, path
length ratio, and small-world Q for the bipartite graphs.20 Their indexes
quantify the small-worldly-ness of an observed bipartite graph, relative
to a baseline random bipartite graph with a constant tie distribution.
Small-worldly graphs are marked by the unusual combination of a short
average path length and a high level of local clustering (Watts 1999).
World polity theory suggests just such a small world: if one IGO is linked
to two other IGOs, those two other IGOs should also be linked, and the
path from any one IGO to another should be short. For instance, in a
completely linked graph with density equal to 1.0, the clustering coefficient
reaches its maximum value of 1.0. Realist approaches also anticipate
clusteringfriends of friends should be friends if conflict and regionali-zation drive the systembut a longer path length since overlap among
IGOs through shared member states should be less common. The clus-
tering ratio is the fraction of the observed clustering over the baseline
clustering for a random bipartite graph. In turn, the path length ratio is
the fraction of the observed average path length over the baseline average
path length for a random bipartite graph (in the random bipartite graph,
the tie distribution is identical to that of the observed graph). The small-
world Q is calculated as clustering length ratio. Below, I reportratio/path
all three indexes, given that both world polity theory and its realist al-
ternatives predict clustering but differ on average path length.
Finally, to assess the regionalization of the world polity, I calculate the
correlation between the observed network of states and a hypothetical
model network in which states are interconnected only within geographic
regions. The regions are the six world macro-regions designated by the
UN: Africa, Asia, Europe, Latin America, Northern America, and Oceania
(United Nations 2005). In the model regional network, all African states
are connected to all other African states but to no other states, all Eu-
ropean states are connected to all other European states but to no other
states, and so on for each region. The Pearson correlation between this
model network and the observed world polity network is calculated using
the Quadratic Assignment Procedure (QAP) as implemented in Ucinet 6,
which randomly reorders, or permutes, the rows and columns of the ob-
served data matrix 2,500 times, recalculating the correlation with the
model matrix for each permutation of the observed matrix. This procedure
19 The random networks are not used to calculate baselines for the preceding measures
since density, centralization, and heterogeneity are also functions of the tie distributions.20 Latapy, Magnien, and Del Vecchio (2006) develop additional clustering measures for
bipartite graphs.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
19/51
American Journal of Sociology
1036
gives a nonparametric test of statistical significance that is appropriatefor nonindependent network data (see Padgett and Ansell [1993] and
Kadushin [1995] for applications of the QAP technique). The Pearson
correlation ranges from 1 to 1.
Throughout the presentation of results, I compare the characteristics
of the observed world polity networks to three ideal-typical networks to
give the reader context for interpretation of the results and to demonstrate
that the changing network structure is not simply a function of the chang-
ing size of the network. The first ideal-typical network is a completely
interconnected and maximally dense networka strong version of the
structural implications of world polity theory. The second ideal-typical
network is a random network of the same size and density as the observed
networkthe results for this random network are used to contextualizethe centralization and coreness analyses. The third ideal-typical network
is a random network of the same size, density, and degree distribution
these results are used in the calculation of the clustering statistics, fol-
lowing the models developed by Newman, Strogatz, and Watts (2001)
and Uzzi and Spiro (2005). As with any ideal-typical or random network
structure, these comparisons are offered as baselines for comparison and
to facilitate interpretation. They underscore the characteristics of the
world polity that are relative to a theoretical model, as well as to size-
and density-independent stochastic models (cf. Anderson, Butts, and Car-
ley 1999).
RESULTS
I present the results in a series of figures. Figures 19 trace change in the
relevant properties of the interstate and inter-IGO networks as they
evolved over 18202000. Figures 1015 are maps that depict the binary,
valued, and domain-specific networks for 2000, in which the nodes in the
network are placed at the location of states capital cities.
Figure 1 verifies the world polity structuration that has been noted in
previous work (Boli and Thomas 1997, 1999). The population of IGOs
grew slowly from one (the Central Commission for the Navigation of the
Rhine, whose members were Baden, Bavaria, France, Germany, Hesse,
and the Netherlands) to 67 in 1930, declined to 63 in 1940, and then grew
rapidly to a total of 330 IGOs by 2000. The 193040 decline in the pop-
ulation of IGOs is consistent with the decline in international trade in
the interwar years (Chase-Dunn, Kawano, and Brewer 2000) and the
steep fall in the formation of INGOs leading up to the Second World
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
20/51
World Polity Social Structure
1037
Fig. 1.Intergovernmental organizations (IGOs) and states in the world polity
War (Boli and Thomas 1997, p. 175).21 But in the later period, the growth
trend in this population of IGOs matches other measures of world polity
formation such as growth in the population of INGOs. This was a massive
increase in the total volume of world polity ties, and there is no doubt
that the amount of world polity activity (or what could be called absolute
population density of the world polity) increased dramatically over the
course of the 20th century (Boli and Thomas 1997; Meyer et al. 1997).
Given that the volume of activity is the network mechanism for diffusion
identified by world polity theory, there is doubtless a very large potential
for policy diffusion in this network.
Figure 1 also shows the widely noted increase in the number of sov-
ereign states in the international system. Through the Concert of Europe
in the early 1800s and the European wars of the later 1800s, this number
grew steadily but very slowly. There was a noticeable increase after the
establishment of the League of Nations in 1919, and then the rate of
growth changed dramatically around 1940. After 1940, the number of
states in the international system grew from 65 to 190 by 2000. The overall
trend, especially the rapid increase since 1940, is consistent with insti-
21 Interestingly, this correspondence between political globalization (as represented by
IGOs and INGOs) and economic globalization (as represented by international trade)offers some support for the argument that international markets demand international
rules (Fligstein 2001; Fligstein and Stone Sweet 2002).
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
21/51
American Journal of Sociology
1038
Fig. 2.Standard deviation, mean, and coefficient of variation in intergovernmental
organization (IGO) memberships.
tutionalist arguments that the world polity (precisely, through the UN
system) legitimizes the state as a form of governance: World society con-
tains much cultural material authoritatively defining the nation-state as
the preferred form of sovereign, rational actor (Meyer et al. 1997, p. 158).
Of course, the overall trend of increasing state formation is consistent
with other arguments as well, such as the dissolution of empires (Wimmer
and Min 2006, p. 871); the network approach taken here cannot adjudicate
these inferences.
Critical for understanding the structural implications of these trends is
the fact that the rapid growth in the number of states and IGOs in the
world politythe growth in absolute population density in the world
politycould have various consequences for the overall structure of the
network. If the new IGOs are universal or nearly universal in their mem-
bership (like the UN, founded in 1945) and if states join IGOs on entry
into the international system (like the example of the island society), then
the network as a whole should become more densely interconnected and
less centralized. States should become more even in their levels of em-
beddedness in the world polity, and states should become more denselyinterconnected through their common IGO memberships. Likewise, IGOs
should become more densely interconnected through their common mem-
ber states. Conversely, if the new IGOs are less like the UN and more
like the EU (founded as the European Economic Community in 1957) or
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
22/51
World Polity Social Structure
1039
Fig. 3.Density of the bipartite, intergovernmental organization (IGO)-by-IGO, andstate-by-state graphs.
if states resist IGO membership, then the network would become increas-
ingly disintegrated and fragmented. Growth in the populations of IGOs
and states does not, by itself, determine the nature of change in network
structure. Instead, because the field of IGOs changes over time, the struc-
ture of the network depends largely on births and deaths in the de-mography of the world polity. These population increases by themselves
are consistent with all the theoretical approaches outlined above.
Figure 2 shows trends in descriptive statistics on IGO involvement
as an attribute of statesto facilitate comparison of these results with
results from previous world polity research. Consistent with world polity
theory, the average number of IGO memberships held by states increased
rapidly, from less than 1.0 through 1860 to 59.1 by 2000 (again, a dramatic
increase in the absolute population density of world polity ties). The stan-
dard deviation also increased but not as quickly as the mean. Conse-
quently, the coefficient of variation decreased from 1.68 in 1820 to .59 by
1910 and further to .37 by 2000. States are growing more similar in level
of involvement in IGOs. These results are consistent with the findings
that inequality in IGO and INGO ties decreased from 1960 to 2000 (Beck-
field 2003), but they extend this work by showing that there has been an
even longer-term trend toward evenness in the depth of states embed-
dedness in IGOs. This growing evenness, like the increases in the pop-
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
23/51
American Journal of Sociology
1040
ulations of IGOs and states, is also consistent with the structural impli-cations of all the theoretical approaches discussed above.
So far, the results shown here replicate previous work using newly
available data compiled by Pevehouse et al. (2004). Next, I turn to the
network analysis, which examines involvement in the world polity as a
relational network rather than an attribute of states. Figure 3 shows trends
in the density (or network density, as opposed to population density or
volume) of three networks: the two-mode network of IGOs and states,
the network of states with overlapping IGO memberships, and the net-
work of IGOs with shared member states. The density of the two-mode
network decreases from 1820 through 1860, increases and remains at a
higher level from 1870 through 1940, then decreases slightly through 2000.
The increase from 1860 to 1870 was driven by the founding of two uni-
versal IGOs in the intervening years: the International Telecommunica-tions Union and the Universal Postal Union. The post-1940 decrease in
the density of the two-mode network is somewhat surprising in light of
the founding of the UN and other universal IGOs after World War II.
Also surprising (relative to the maximally dense ideal-typical network
drawn from world polity theory) is the sparseness of the bipartite network
throughout the period: in every decade, fewer than 40% of the possible
ties between IGOs and states are realized.
The density of the one-mode interstate network is more consistent with
world polity theory than with the alternative approaches: very low pro-
portions of possible ties among states are realized until the founding of
universal IGOs after 1860, and then the density increases rapidly and to
a very high level by the turn of the century. By 1900, fully 88% of possible
ties among states are realized. With the exception of two slight dips, the
density of the interstate network continues to increase through 2000 to
97% of possible ties realized. This means that nearly every state holds at
least one IGO membership in common with every other state (unsur-
prising, given the near-universal membership of the UN). If density eases
institutional diffusion and enhances normative emulation (DiMaggio and
Powell 1983, p. 152; Henisz et al. 2005, p. 876), world polity models
circulate easily among states in the international system. Of course, it is
important to emphasize that density is calculated for binary networks, so
that information on the strength of ties is reduced to a binary indication
of the presence/absence of at least one tie. The ultimate significance of a
single tie would then depend on the significance of the one organization
that formed the tie.Turning from ties among states to ties among IGOs, we find a trend
in density that is almost the mirror image of the trend in the density of
interstate ties: IGOs are very densely interconnected through their mem-
ber states at first, but this density slowly declines through 1940, then
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
24/51
World Polity Social Structure
1041
Fig. 4.Centralization in the intergovernmental organization (IGO) and interstate net-works.
declines more rapidly through 1990, when it appears to level off at just
over 50% of possible inter-IGO ties observed. While this remains a fairly
dense network in a static sense, the pattern of change is more consistent
with the field-theoretic economic-sociology and international-relations ap-
proaches than with world polity theory. The decreasing density of the
IGO network suggests a more fragmented world polity, with fewer con-
nections among organizations. This means that the pattern of IGO ties
contrasts starkly against the number of IGO ties: while states are growing
more even in the number of IGOs they belong to, they increasingly belong
to different IGOs. In contrast to a world of UNs, WTOs, International
Labor Organizations, and World Banks, the decreasing density of the
IGO network results from a world of EUs, North American Free Trade
Agreements, Mercosurs, and Association of South East Asian Nations.
Moving from the density of interconnections in the world polity to its
potentially stratified structure, figure 4 shows the trends in centralization,
a measure of structural heterogeneity that assesses how closely a network
corresponds to a star shape with one central node and many peripheral
nodes. Again, we find opposite trends for the inter-IGO and interstate
networks. Centralization of the interstate network drops sharply from1840 to a low level in 1910 and decreases even further through 2000. In
2000, the centralization of the interstate network fell to .07, very close to
the perfect structural equality of a circle-shaped network in which all
nodes are interconnected (and actually less than .11, which is the cen-
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
25/51
American Journal of Sociology
1042
Fig. 5.Heterogeneity in coreness of states
tralization for a random network in which size and density, but not the
tie distribution, are held equal to that of the 2000 world polity network).
Decreasing centralization of the interstate network is consistent with the
image drawn from world polity theory of an increasingly even field of
association and inconsistent with the world-systems approach.
In contrast, the centralization of the IGO network increases throughout
the period. The IGO network exhibits no structural heterogeneity until
1880, but the level of centralization then increases slowly through 1940,
when the network is 16% as centralized as possible given its size. Cen-
tralization then rises more steeply through 1970 and ultimately levels off
near 48% (much greater than the 17% centralization in the equivalent
size and density random graph). This suggests that the IGO network is
increasingly star shaped, with central IGOs that share many member
states in common with each other and peripheral IGOs that share member
states in common with central IGOs but not with other peripheral IGOs.
This increasing centralization of the IGO network is more consistent with
the neorealist and world-systems images of a world polity structured by
international competition and conflict.
Centralization, like density, is calculated for binary networks, whichmeasure only the presence or absence of ties and waste information on
the strength of ties among IGOs and states. Centralization also has limited
utility for measuring structural heterogeneity in these networks because
centralization assesses the resemblance of a network to an ideal-typical
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
26/51
World Polity Social Structure
1043
Fig. 6.Heterogeneity in coreness of intergovernmental organizations (IGOs)
network with only one central node. Borgatti and Everett (1999, p. 376)
generalize this restricted sense of centralization and define a core/periphery
structure as one in which the network . . . consists of just one group to
which all actors belong to a greater or lesser extent, when a center and
periphery can be defined. That is, core/periphery networks can be char-
acterized as having a multicentric core of nodes that are tightly inter-
connected, with a surrounding periphery of less interconnected nodes.22
The overall level of structural heterogeneity in a core/periphery network
can then be measured as dispersion in the closeness of the nodes to the
core. Figure 5 shows trends in the Gini coefficient, a common measure
of dispersion, calculated for the binary and valued interstate networks.
Figure 5 confirms that the binary network understates the level of
structural heterogeneity among states: from 1860 onward, the Gini co-
efficient is always lower for the binary network than for the valued net-
work. For instance, in 2000, the Gini coefficient for coreness scores in the
valued network is .144, while in the binary network it is .001 (for random
graphs of the same size and density, these quantities are .063 and 0).
Nevertheless, the trends are similar: in both state networks, structural
heterogeneity declines steeply after 1960 to a very low level by 2000. For
22 Correlations between the observed valued networks and idealized core-peripherystructures (Borgatti and Everett 1999) are substantial. For all years of observation,
the correlation coefficients for the valued networks surpass .80.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
27/51
American Journal of Sociology
1044
Fig. 7.Cohesive world polity? Average path length, diameter, and connectivity
comparison, the level of inequality in world income has been estimated
at a Gini coefficient of .543 (Firebaugh 1999, p. 1613). The level and trend
are consistent with world polity theory: all states are nearly equal in
position within the world polity.
Moving from the interstate network to the inter-IGO network, figure6 shows the trends in this measure of structural heterogeneity for the
binary and valued networks of IGOs. The key distinction between figures
5 and 6 is that figure 5 shows results for the state mode, whereas figure
6 shows results for the IGO mode. And, once again, the IGO structure
appears to be at odds with the state structure: in both the binary and
valued IGO networks, structural heterogeneity trends upward after 1870,
with notable growth in the rate of increase during 194070. The level of
heterogeneity is, predictably, higher in the valued network (Ginip .578
in 2000) than in the binary network ( in 2000). Both coeffi-Ginip .197
cients are much larger than those for random graphs of the same size and
density (.087 and .001, respectively). For the IGO network, the level and
trend are consistent with the global neorealist approach: central IGOs
like the UN and WTO are much closer to the structural core of the world
polity than peripheral IGOs like the Baltic Council and the Central Asian
Economic Community. The implications for diffusion are that policy
scripts diffuse much more easily through the core of the world polity,
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
28/51
World Polity Social Structure
1045
creating the potential for a fragmented and bumpy pattern of policydiffusion.
Figure 7 shows the cohesion indexes: average path length, diameter,
and minimum connectivity. These measures begin the assessment of net-
work topology: recall that several theoretical approaches to political glob-
alization, including world polity theory, world-systems theory, and neo-
liberalism in international relations, all imply a small, cohesive world
polity, with very little network distance separating IGOs and states. The
average path length (geodesic distance) of the bipartite graph increased
at first through the mid-19th century, decreased again briefly, then in-
creased fairly consistently through the 20th century. This is preliminary
evidence that the world polity looks less, not more, like a small world,
even since the 1940s and the emergence of the UN system. It is worth
noting that the average path length for a random graph of identical sizeand density is 2.298, less than the 2.678 average path length observed in
the world polity in 2000. This decrease in the cohesion of the world polity
is also reflected in the graphs diameter, which grows to five links long
during 195090 (it decreases again to four in the 2000 network, which is
equal to the diameter of a random graph of the same size and density).
The results for connectivity show a network that is disconnected through
1950, after which the connectivity increases. Still, in 2000, only three nodes
would have to be removed to disconnect the graph. For comparison, a
random graph of the same size and density shows a connectivity of 21
nodes. The results for average path length, diameter, and connectivity,
thus, suggest that the IGO network more closely resembles the fragmented
world of the realist approaches than it does the cohesive world of world
polity theory. It is no more cohesive than a random network of the same
size.
Figure 8 extends the analysis of network topology from general size
and cohesion to small-world structure (Watts 1999). The world polity has
also become less small worldly over time. The clustering ratio decreases
from 1.0 (meaning that the graph is exactly as clustered as a baseline
random graph with equivalent size, density, and tie distribution) in the
early years to around .94 for 18901940, then decreases more rapidly to
.84 by 2000. What this means is that the bipartite graph is actually less
clustered than would be expected given that states are automatically clus-
tered within IGOs in this bipartite graph. That is, there is less between-
IGO clustering than would be expected for a random bipartite graph with
this tie distribution. What clustering does exist remains within IGOs.Interestingly, this decreasing clustering coincides with an increasing path
length ratio, meaning that the states and IGOs in the bipartite graph are
more distant from each other than would be expected given the bipartite
structure of the graph. All this results in a small-world Q that follows a
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
29/51
American Journal of Sociology
1046
Fig. 8.Small world polity? Clustering ratio, path length (PL) ratio, and small-world Q
generally negative trend throughout the period and declines more rapidly
after the 1940s. Since this is precisely the period when the development
of the UN system could have been expected to produce more crosscutting
ties among states, it would seem that the regionalization of the system in
the later period outweighed this development in structural implications.
The results so far are consistent with the regionalization of the world
polity, but a more direct test is required, in light of the divergent theoretical
implications developed above for the degree of regionalization of the world
polity. To what degree does the regionalization of the world polity shape
its structure? Figure 9 shows trends in the correlation between the ob-
served interstate networks and a model regional network (in which states
share IGO ties only to other states in their geographic region, as defined
above).23 For the binary network, regionalization increases until 1870,
23 To assess the sensitivity of these results to the UNs definition of region, I reestimated
the correlations using the alternative, more culturally oriented scheme of Huntington(1996), as operationalized by Henderson and Tucker (2001). The nine civilizations
identified by Huntington correspond fairly closely to a strictly geographical grouping,
except that some North African and Middle Eastern countries are classified as Islamic,
North American countries are classified with Western European states as Western, andthe Asian countries are divided among the Buddhist, Japanese, and Sinic civilizations.
Using this alternative semiregional classification, the results are consistent with those
shown: the association between region and IGO ties increases through 1900, decreasesthrough 1920, increases through 1940, falls off sharply through 1970, then increases
again (to the level of .325). Details are available from the author.
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
30/51
World Polity Social Structure
1047
Fig. 9.Correlations between observed networks and regional model networks
decreases until 1950, and stabilizes at a very low level through 2000
(Pearsons ). For the valued network, the trend is rather different:rp .009
regionalization increases sharply and peaks in 1930, decreases dramati-
cally after 1940 because of the founding of the UN, and increases again
after 1960 to a correlation of .365 in 2000. The prewar peak of region-
alization contradicts the argument that universal IGOs dominated the
field in the first half of the 20th century and fostered regional organizations
only after the war (Boli and Thomas 1999, p. 31). The overall pattern of
findings, especially the contrast between the results for the binary versus
valued networks, suggests that sharing a geographic region may not pre-
dict the presence or absence of ties among states, but it does predict
stronger ties among states. States do have some connections that span
regional boundaries, but connections among states are strengthened
within regions. More important than the static patterning of interstate
ties is the direction of change: the regionalization of the world polity has
actually increased since 1960, and world polity ties have become more
rather than less bound to place. This is consistent with Nierops (1989)
finding that the regional clustering of states grew during the postwar
period through 1980. Taken together, the results shown in figures 8 and9 indicate that the world polity more closely resembles a regionalized
world than a singular small world.
Before turning to a more detailed examination of the contemporary
structure of the world polity, I note two aspects of its long-term evolution
-
7/30/2019 BECKFIELD REdes de politica mundial.pdf
31/51
American Journal of Sociology
1048
in the context of some of the key historical changes in the geopolitics ofthe period. For nearly all the structural measures, stasis, or at least trend-
less fluctuation, characterizes most of the period before the late 19th cen-
tury. This relative lack of change in the IGO system may reflect the
conclusion of the nationalization period of state formation (Tilly 1990, p.
185), when national interests surpassed dynastic ones within European
states. The static structure of the world polity during this period of modern
state formation reinforces the link between national sovereignty and the
institutionalization of the world polity (Meyer et al. 1997) and accords
with the identification of the late 19th and early 20th centuries as the
crucial period of world polity formation (Boli and Thomas 1999). But the
truly transformative geopolitical moment that ripples through the struc-
ture of the world polity is the regionalization of the world polity that
occurs largely after the establishment of the UN system. After 1945, theworld polity grows more disintegrated, more centralized, more structurally
uneven, and more fragmented by increasingly regionalized IGOs. It re-
sembles less and less one small worldwhen the structure of the inter-
IGO network is accounted for.
Conversely, the generally placeless, even, flat structure of the interstate
network is shown in figure 10, which is a mapping of the network data
to geographical coordinates (i.e., longitude and latitude of capital cities)
for 2000. As will become clear with the comparison to figure 11, the key
distinguishing feature of this map is that it does not display the strength
of the IGO ties connecting states: on this map, states are linked if they
have at least one IGO membership in common. That is, all IGOs are
treated equivalently, and the structure of the IGO network itself is thus
ignored. The resulting network of states exhibits a notably even, decen-
tralized, flat structure. The ties (lines) are so dense that they cannot be
distinguished, as every state is tied to nearly every other state through
their common IGO memberships.24 Indeed, the locations of capital cities
24 Supplemental analysis (available on request) of the IGO network shows that the
central, nonregional IGOs include highly visible, prominent organizations such as theUN, the WTO, UNESCO, Interpol, the World Bank, and the IMF. Their visibility
and influence accord with their central position within the IGO network. Another
prominent IGO, the OECD, occupies a potentially strategic brokerage position (Burt
1992, 2004) between the European cluster and the Americas cluster. Following Burt(1992, 2004), this suggests that the OECD may act as a bridge between these regions
and thus more effectively diffuse policy scripts among them. While the influence of
the OECDs structural position on its ability to diffuse policy scripts is, of course,
somewhat speculative, this speculation provides an example of the kind of hypothesesthat can be generated and tested by conceptualizing and observing the world polity
as a network. Viewing the world polity as a network also genera